Pattern-based exceptions in .gitignore not working

[swisspol]

Example case:

*.data
!dont_ignore/*.data

Then add a couple files in the workdir: foo.data and dont_ignore/bar.data.

git_diff_index_to_workdir() will correctly ignore foo.data but incorrectly also ignore dont_ignore/bar.data.

Note that the .gitignore file is working as expected when testing with git status --ignored.

[swisspol]

Doing a bit more testing, this exception works: !test.data.

[carlosmn]

It looks like the *.data rule satisfies the restriction that negation rules must negate an existing pattern.

void test_status_ignore__deeper(void)
{
   int ignored;

    g_repo = cl_git_sandbox_init("empty_standard_repo");

    cl_git_mkfile("empty_standard_repo/.gitignore",
          "*.data\n"
          "!dont_ignore/*.data\n");

    cl_git_pass(p_mkdir("empty_standard_repo/dont_ignore", 0777));
    cl_git_mkfile("empty_standard_repo/foo.data", "");
    cl_git_mkfile("empty_standard_repo/bar.data", "");
    cl_git_mkfile("empty_standard_repo/dont_ignore/foo.data", "");
    cl_git_mkfile("empty_standard_repo/dont_ignore/bar.data", "");

    cl_git_pass(git_ignore_path_is_ignored(&ignored, g_repo, "foo.data"));
    cl_assert_equal_i(1, ignored);
    cl_git_pass(git_ignore_path_is_ignored(&ignored, g_repo, "bar.data"));
    cl_assert_equal_i(1, ignored);

    cl_git_pass(git_ignore_path_is_ignored(&ignored, g_repo, "dont_ignore/foo.data"));
    cl_assert_equal_i(0, ignored);
    cl_git_pass(git_ignore_path_is_ignored(&ignored, g_repo, "dont_ignore/bar.data"));
    cl_assert_equal_i(0, ignored);
}

[fogaztuc]

Just found this issue as well.

[overboen]

I believe that there is an issue in the Atom tree-view project that is caused by this via Atom's git-utils' project's dependency on this project. It looks like the project is using v0.24.3 of libgit2.

atom/tree-view#902

[segevfiner]

Not sure if this is the same, but in the CPython repository they have this:

/python
!/Python/

The "Python" directory is ignored by libgit2 but not by Git.

Tested on Windows.

[pks-t]

@segevfiner I guess your issue is different, as it is caused by issues with case-insensitivity. It would be very nice if you could create a new issue describing the problem.